Hybrid Computational Strategy for Predicting Complex Ligand-Metal Architectures.

Area of Science:

• Computational chemistry
• Materials science
• Chemical informatics

Background:

• Predicting metal-ligand coordination is crucial for designing metal complexes and catalysts.
• Ligands can exhibit numerous coordination modes, posing challenges for chemists.
• Existing methods struggle with complex coordination patterns.

Purpose of the Study:

• To develop a computational approach for predicting complex metal-ligand coordination patterns.
• To create a versatile model applicable to a wide range of ligands and metals.
• To provide an accessible tool for chemists.

Main Methods:

• A hybrid computational approach combining Machine Learning (ML) and knowledge-based rules.
• Training an ML algorithm on data from the Cambridge Structural Database (CSD).
• Developing a predictive model for coordination patterns.

Main Results:

• The ML model successfully predicts complex coordination patterns for various ligands and metals.
• The approach handles diverse ligand types, including hemilabile, haptic, and high-denticity ligands.
• The model is effective across different metal oxidation states.

Conclusions:

• The developed hybrid ML approach offers a robust solution for predicting metal-ligand coordination.
• The tool enhances the rational design of metal complexes and catalysts.
• The algorithm is available via RDKit and a public web portal.